Privacy-Aware Customer Segmentation Using a Distributed Graph-Based Attribute Projection Framework

Customer segmentation plays a vital role in Business Intelligence (BI) by enabling organizations to understand customer behavior, enhance personalization, and support informed decision-making. Conventional segmentation approaches, including K-Means clustering, hierarchical methods, and hybrid deep learning models, often face limitations when handling high-dimensional customer data and typically lack built-in mechanisms to address privacy concerns. As customer analytics increasingly relies on sensitive personal information, these limitations pose significant challenges for responsible data-driven applications. To overcome these issues, this study introduces a Distributed Graph-Based Attribute Projection Framework (GAPF) for privacy-aware customer segmentation. The key novelty of the proposed framework lies in its ability to minimize sensitive attribute exposure while preserving meaningful relational patterns among customers through graph-based representations. GAPF employs a distributed processing pipeline that integrates attribute projection to reduce identifiability, heuristic-driven customer similarity graph construction, graph convolutional network (GCN)–based feature learning, and community detection for final segmentation. The framework is implemented using Python, NetworkX, and PyTorch Geometric and evaluated on the Mall Customers dataset and large-scale anonymized synthetic data to assess scalability. Experimental results demonstrate that GAPF achieves superior segmentation performance, with an accuracy of 98%, precision of 92.5%, recall of 94.0%, and an F1-score of 93.2%, while also exhibiting efficient execution and reduced privacy risk. These findings confirm GAPF as a robust and practical solution for privacy-aware BI applications.